Understanding feeding patterns in growing pigs by modelling growth and motivation

Applied Animal Behaviour Science 171 (2015) 69–80

Contents lists available at ScienceDirect

Applied Animal Behaviour Science journal homepage: www.elsevier.com/locate/applanim

Understanding feeding patterns in growing pigs by modelling growth and motivation Iris J.M.M. Boumans a,∗ , Eddie A.M. Bokkers a , Gert Jan Hofstede b , Imke J.M. de Boer a a b

Animal Production Systems group, Wageningen University, P.O. Box 338, 6700, AH Wageningen, the Netherlands Information Technology group, Wageningen University, P.O. Box 8130, 6700, EW Wageningen, the Netherlands

a r t i c l e

i n f o

Article history: Received 23 February 2015 Received in revised form 18 June 2015 Accepted 10 August 2015 Available online 22 August 2015 Keywords: Pig feeding behaviour Meal patterns Motivation Growth Agent-based model

a b s t r a c t Feeding is an essential behaviour for body maintenance in pigs and closely related to their growth and productivity performance. Mechanisms underlying feeding behaviour in pigs are still unclear. Understanding these mechanisms can provide valuable insights into the complex interactions among various factors affecting feeding behaviour and help to improve growth and productivity of pigs. The aim of this study was to increase our understanding of internal causation and development of short-term feeding patterns in a pig, and the relation between feeding patterns and productivity of a pig during the growth period. We developed a mechanistic simulation model that represents an individually housed growing pig. The model integrates knowledge from physiology and ethology, and combines growth with a behavioural decision model based on motivation. Combining growth with behaviour allowed exploring the development of a pig over time, in particular the causation of growth and feeding patterns over a 24 h period and during the entire growing period. Physiological factors, affected by pig and feed characteristics, are important internal factors controlling feeding behaviour. Model output included short-term feeding behaviours in pigs (e.g. meal size, meal frequency and meal duration), and growth characteristics (e.g. energy use, body weight gain). The model yielded feeding patterns that were validated against empirical data. This modelling study provided insight in how growth and motivation explain the development of feeding patterns of an individually housed pig over time. Pig and feed characteristics affected the motivation to reach a desired level of daily feed intake. Without feeding restrictions, pigs adapted feeding patterns to reach this daily feed intake without affecting growth. The developed model is suitable to further study mechanisms underlying feeding behaviour and performance of group-housed pigs. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Feeding is an essential behaviour for body maintenance in pigs and closely related to their growth and productivity performance (Nyachoti et al., 2004). The amount of feed consumed directly affects the level of nutrient intake (Nyachoti et al., 2004), whereas the distribution of meals over the day affects the utilisation of nutrients (Batterham and Bayley, 1989; De Haer and De Vries, 1993b). Improving feed intake to optimize growth and productivity is still a major goal in pig production (e.g. Kanis, 1990; Rauw et al., 2006; Thingnes et al., 2012). Feed intake results from the complex interaction of a number of factors. Multiple external factors are known to affect feed

∗ Corresponding author. Tel.: +31 317 48 38 82. E-mail addresses: [email protected] (I.J.M.M. Boumans), [email protected] (E.A.M. Bokkers), [email protected] (G.J. Hofstede), [email protected] (I.J.M. de Boer). http://dx.doi.org/10.1016/j.applanim.2015.08.013 0168-1591/© 2015 Elsevier B.V. All rights reserved.

intake in pigs, such as dietary composition (Brouns et al., 1994), ambient temperature (Quiniou et al., 2000), group housing (Bornett et al., 2000), and environmental stimuli (Beattie et al., 2000). Feed intake, however, also differs among individuals kept under similar conditions (Nielsen, 1999; Bornett et al., 2000), due to a number of internal factors, e.g. genotype and sex (De Haer and De Vries, 1993a), body weight (Quiniou et al., 2000), physiological stage of development (NRC, 2012), and health (Williams et al., 1997). Although much is known about the effects of the aforementioned external and internal factors on feeding, it is still unclear which underlying mechanisms are responsible for the observed feeding patterns, such as daily feed intake, meal frequency and meal size. Understanding the underlying mechanisms of feeding patterns in pigs can provide valuable insights into the complex interactions between various factors and help to improve the growth and productivity of pigs. Several mechanistic models have been developed to simulate feed intake and daily growth in pigs. These models include, for example, feed composition and nutrient partitioning (e.g. De Lange,

70

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

1995; De Lange et al., 2003), environmental aspects such as stocking density and temperature (e.g. Yoosuk et al., 2011), or animal characteristics such as initial body weight, growth potential, and ability to cope with social stressors (e.g. Wellock et al., 2004). These models, however, do not include short-term feeding patterns, such as meal frequency, meal size, meal duration, and between-meal intervals. Models that do include short-term pig feeding patterns (e.g. Morgan et al., 2000; Lewis and McGlone, 2008) are empirical (regression) equation models, which do not model mechanistic effects on feeding patterns, which are essential to gain more insight into mechanisms underlying feeding patterns. This study is the first step of a larger research project, in which we want to gain more understanding of feeding patterns in pigs and the role of interactions among individuals in one pen. We first want to understand internal processes controlling feeding behaviour, before we include social interactions among pigs. Therefore, the aim of this study was to create a model that would increase our understanding of internal causation and development of shortterm feeding patterns in a pig and the relation between feeding patterns and productivity of a pig during the growth period. In the case of an individual pig housed in a stable environment with ad libitum feed, we hypothesized that feeding patterns emerge from metabolic processes and ethological processes. Our model, therefore, included a constant interaction between growth and behaviour, caused by motivation, and affected by feed and pig characteristics. Combining growth with motivation allows studying the development of growth and feeding patterns in pigs over a 24 h period and during the whole growing period. In this study, we used an agent-based model, which assists with the understanding of emergent behaviour resulting from interactions among individuals in a specific environment (Asher et al., 2009; Railsback and Grimm, 2012). We selected this type of model here because it meets the goal of the larger research project. Similarly, no attempt was made to simplify the processes included in the model, because these might become important when more than one agent is introduced in the model. For model validation, results were compared with empirical data from literature. In addition, a local sensitivity analysis (Railsback and Grimm, 2012) was performed to assess how individual parameters in the reference settings affected model output and which parameters and conditions were important for the model results.

2. Model description 2.1. Process overview The simulation model was constructed and implemented in the computer program Netlogo version 5.0.3 (Wilensky, 1999). The model represented one individually housed growing pig in a conventional pen, with ad libitum access to water and commercial dry feed, which met the nutrient requirements for maintenance and growth of the pig. Pig behaviour emerged from the interaction of two sub-models (Fig. 1). In the sub-model called ‘Motivational decision-making’, the pig updated its four motivational states: feeding motivation, resting motivation, drinking motivation, and exploring motivation. These motivational states simultaneously affected its decision per time step to perform a specific behaviour, or (when motivational states were not high enough) remaining lying or standing. The chosen behaviour affected the energy use and feed intake of the pig, which subsequently changed its nutrient balance and growth, modelled in the second sub-model called ‘Growth’. The outcome of the sub-model Growth was input again into the first sub-model. This interaction was modelled in time steps of one minute, adding up to days, as a continuous cycle over

Table 1 Input and output variables in the model. Input variables

Output variables

Pig characteristics Sex Initial body weight (kg) Initial body protein weight (kg) Mean body protein deposition (g) Minimum body lipid to body protein ratio

Feeding behaviour Feed intake (g/day) Meal frequency (no/day) Meal size (g/meal) Meal duration (min/meal) Feeding rate (g/min) Meal interval time (min) Feeding time (min/day)

Feed characteristics Digestible energy content of the diet (kJ/g) Dietary protein content (g/kg) Apparent protein availability Dietary amino acid content (g/kg) Apparent amino acid availabilities Diet digestion duration (min) Diet palatability

Growth Body weight gain (g/day) Energy use (kJ/day) Body weight (kg) Empty body weight (kg)

the entire growing period of 120 days. Each day (1440 minutes, 24 h) was modelled with a light period from 08:00 till 19:00 h, and a dark period in the remaining hours. This reflected a common light regime and complied with the EU legislation requirements of a minimum of 8 h of light per day. Model output variables were the feeding and growth patterns of one pig (Table 1). Feeding patterns were calculated based on the behavioural decisions of the pig at each minute during the day. A meal started at a time step when the pig performed feeding behaviour and ended at the time step when it stopped feeding. Meal duration was calculated as the number of successive minutes that the pig was feeding for. Meal size was the multiplication of meal duration with feeding rate per minute. On this basis, average interval time between meals, average meal frequency, total feeding time and total feed intake were calculated per day. Model input variables on pig and feed characteristics are given in Table 1. Feed had a value for energy content, protein content and availability, digestibility, and palatability. The modelled pig was either a gilt, barrow or male, with a genotype effect based on a mean body protein deposition curve and a minimum body lipid to body protein ratio. Simulations started with a pig at 70 days of age, weighing approximately 27 kg. 2.2. Sub-model: Motivational decision-making The decision-making process of a pig to perform a given behaviour in the model was based on motivations. To let the pig autonomously decide on when and how to feed, a comprehensive motivational system for the causation of feeding motivation was included. In addition, to allow the pig to change its behaviour during a 24 h period and to model the effect on metabolic energy use, (less comprehensive) motivational systems for resting, drinking, and exploring were included. 2.3. Causation of feeding motivation Feeding motivation and feedback mechanisms can be used to explain the interaction and integration of factors that control feeding behaviour (Day et al., 1998). Causation of feeding motivation in pigs was modelled analogous to work by Sauvant et al. (1996) on feeding decisions in sheep. Feeding motivation was a balance between feeding drive and satiation (Fig. 2). Feeding drive increased feeding motivation, whereas satiation reduced feeding motivation: the pig was motivated to feed when its feeding drive was higher than its satiation. Two types of internal factors controlling feeding behaviour are often described: metabolic-homeostatic and cognitive-hedonic factors (Berthoud and Morrison, 2008; Johnson, 2013). Metabolic-homeostatic factors concern energy and

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

71

New me step (minute)

Sub-model Growth

Feed characteriscs

Sub-model Movaonal decision-making

Update nutrient balance & grow

Pig characteriscs

Update movaons & choose behaviour

Behaviour

Fig. 1. Overview of general processes in the model. Pig behaviour is caused by motivation, which is affected by pig characteristics, feed characteristics and the sub-model Growth. Arrows indicate causal relationships in the model.

nutrient levels within an animal (Johnson, 2013) and were, in this study, included as instantaneous factors (e.g. instant energy balance) affecting satiation, or daily factors (e.g. daily energy balance) affecting feeding drive. Cognitive-hedonic factors concern liking and wanting food and environmental cues (Berthoud and Morrison, 2008), and were here included as diet palatability and diurnal rhythm, affecting the feeding drive. Technical details of the causation of feeding motivation in the model are described in appendix A.1. Feeding motivation can increase or decrease via feedback mechanisms. These mechanisms prevent unnecessary switching between behaviours and ensure that behaviours last long enough to reach the functional goal of a behaviour (Mason and Bateson, 2009). Positive feedback increases motivation when performing the related behaviour to maintain the performance of that behaviour, and it prevents behaviours with high priorities from inhibiting each other (Mason and Bateson, 2009). Positive feedback

(Reinforcement feeding motivation) was included in the present model by increasing feeding motivation with 0.40 as long as satiation was below 0.95 during feeding. The value 0.40 was chosen after calibration of the model on expected feeding patterns, the value 0.95 was chosen with the assumption that reinforcement of feeding motivation continues until a high satiation threshold. Negative feedback reduces motivation and can terminate a behaviour (Mason and Bateson, 2009). Negative feedback occurred in the present model autonomously by a decreased feeding drive and increased satiation after feed intake. Performance of an alternative behaviour and passage of time can also function as negative feedback mechanisms in motivational models (Hogan, 1997). We assumed that for a pig with ad libitum feed access, these feedback mechanisms would not likely reduce feeding motivation, because internal nutritional requirements will increase over time. The duration of a behaviour depended on the level of the involved motivation and the levels of other motivations, which were updated

Fig. 2. Schematic overview of the causation of feeding motivation in the sub-model Motivational decision-making. Feed characteristics, pig characteristics and sub-model Growth affect the causation of feeding motivation. Arrows indicate causal relationships in the model. Numbers refer to equations which are described in appendix A.1.

72

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

Table 2 Input parameters and fixed values in the model for causation of motivations to drink, explore or lie down. Variables

Values

Threshold variables Resting 0.25 Light period (08:00 to 19:00 h) 0.20 Dark period (19:00 to 08:00 h) Energy variables (change per minute) Energy level increase 0.033 when behaviour not performed 0.021 Energy level decrease when behaviour performed

Drinking 0.25

Exploring 0.25

0.30

0.30

0.001

0.007

0.279

0.258

at each time step. When a behaviour was intervened, the behaviour could be continued in the next time step if the motivation was still high enough.

Table 3 Input parameters and (initial) values in the model for body maintenance and growth. Parameter Growth pigs Initial body protein weight (kg) a,b Initial body weight (kg)b Minimum body lipid to body protein ratioa Mean body protein depositionb

Diet composition and digestibilityc DE content diet (kJ/g) Dietary amino acid content Lysine (g/kg) Dietary amino acid content Methionine (g/kg) Dietary amino acid content Methionine + Cystine (g/kg) Dietary amino acid content Threonine (g/kg) Dietary amino acid content Tryptophan (g/kg) Dietary amino acid content Isoleucine (g/kg) Dietary protein content (g/kg) Apparent amino acid and protein availabilitiesd

Value 4 27 1 137, 133, 151 g/day for gilts, barrows and males respectively 14.2 11 3 6 6 2 5 132 0.82

a

De Lange, 1995. NRC, 2012. Values were formulated to meet dietary requirements of growing pigs with body weight 50-75 kg and mean body protein deposition rate of 155 g/day based on apparent ileal digestible basis of the diet as described in Table 16- 3A in NRC, 2012. d Based on average ileal digestibility of amino acids in pig diets from Sauer and Just cited in Moughan (1995). b

2.4. Causation of other motivations Besides feeding motivation, the pig in the model had the motivation to rest, explore and drink. These other motivations for behaviours often accompany feeding motivation. An increased feeding motivation will increase exploratory behaviour (rooting) and decrease lying behaviour (Day et al., 1995), whereas drinking behaviour often occurs around feeding behaviour (Bigelow and Houpt, 1988). Furthermore, these other behaviours can affect the energy use of a pig, which in turn can affect its growth and feeding motivation. Therefore, although less detailed as feeding motivation, motivations to rest, explore and drink were also included in the model. Causation of these behavioural motivations was based on motivational theory that includes an energy variable and threshold variable (Hogan, 1997). The energy variable represents an internal build-up of energy (drive) to perform a certain behaviour. The threshold variable limits the performance of the behaviour by requiring a minimum level of energy. In case the energy variable for resting behaviour exceeded its threshold level, for example, the pig became motivated to perform resting behaviour. The value of energy variables increased gradually each minute and decreased when the related behaviour was performed, as mentioned in the Lorenz model (described in Mason and Bateson, 2009). The value of threshold variables depended on the related behaviour and the time of day (Table 2). The fixed values for increase and decrease of energy variables each minute (Table 2) were calibrated to obtain common behavioural time budgets of pigs. At the start of the model, initial values of energy variables were set to a random value based on a normal distribution with a mean corresponding to the threshold value and a 10% standard-deviation. Growing pigs in barren housing on average lie down for over 80% of their 24 h time budget, whereas feeding behaviour occupies approximately 10% of their time budget, and standing and drinking behaviours occupy approximately 8% of their time (Gonyou et al., 1992). Pigs feed, drink and stand mostly during their active (=light) period (Gonyou et al., 1992). In this period, barren housed growing pigs spend over 70% of their time lying, approximately 10% feeding and drinking and 15% exploring (Bolhuis et al., 2005). Energy variables were modelled in such a way that absence of motivations could occur. Without sufficient motivation, a pig remained in the position (standing or lying) of the last behaviour. This prevented that the time budget of pigs in the model was fixed. Furthermore, this could represent a possible mechanism underlying the observed high amount of lying behaviour of pigs in conventional housing systems, which is that lying behaviour occurs due to resting motivation but is extended due to absence of stimuli.

c

2.5. Behavioural decision-making Behavioural decision-making is modelled by the state-space approach (McFarland and Sibly, 1975). This approach implies that the behaviour related to the highest motivation will be performed until another motivation becomes higher or a certain threshold is reached. A pig in the model makes a decision each minute to act on the highest motivation (when above zero). When no motivation is above zero, the pig performs no behaviour, but remains in the position of its last behaviour (standing or lying). When a pig chooses to feed, it first determines its feeding rate, which is based on a preferred feeding rate affected by palatability of the feed and satiation of the pig (technical details in appendix A.2). 2.6. Sub-model: Growth The sub-model Growth was based on the mechanistic pig growth model by de Lange (1995) on nutritional partitioning and growth in the pig. The time step in this model was converted from days to minutes and growth was included as a function of digested feed per minute, which allowed growth as a factor to affect feeding motivation during the day. Fig. 3 shows how absorption and use of nutrients from digested feed results in a new body composition and body weight (BW) each minute. Switching to a time step of one minute allowed for negative growth to occur if required nutrients were extracted from the body when nutrient absorption from digested feed was insufficient for body maintenance. Input parameters for this sub-model are listed in Table 3, growth output parameters are listed in Table 1. The relatively simple mechanistic model by de Lange (1995) contains the essential functions to predict growth (Black and Lange, 1995; De Lange, 1995), but has inadequate assumptions on feed intake, energy use, and the effect of genotype on growth for use in a dynamic simulation model (Black, 1995). Equations for calculating feed intake and energy requirements in the model of de Lange (1995) were replaced by mechanistic results (feed intake, digested feed and energy use) of the sub-model Motivational decision-making. For the effect of genotype on growth more

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

73

New me step (minute) Sub-model Growth

Feed characteriscs Energy value Protein (P) and amino acid (AA) value

Energy absorpon

P & AA absorpon

Pig characteriscs Energy, P & AA for maintenance

Body weight & composion

Sub-model Movaonal decision-making

Remaining Energy, P & AA

Sex

Potenal P deposion

Mean body P deposion

Potenal L deposion

Actual growth

Minimum body P to body lipid (L) rao

Behaviour

Fig. 3. Schematic overview of the sub-model Growth (based on De Lange, 1995), and the interaction with the sub-model Motivational decision-making, behaviour, pig characteristics and feed characteristics each minute. Arrows indicate causal relationships in the model.

3. Results 3.1. Empirical validation of simulated feeding patterns

30

Feeding rate (g/min)

dynamic processes on protein deposition (PD) were included (see appendix A.3 for technical details).

25 20 15 10 5 130

120

110

100

90

80

70

60

50

40

30

0

To validate the model, model output was compared with the empirical study of Bigelow and Houpt (1988) on feeding patterns of pigs. Bigelow and Houpt (1988) studied the feeding patterns of six individually housed female Yorkshire pigs, increasing in BW from 10 to 130 kg. Pigs were kept at a constant temperature of 22–23 ◦ C, and obtained a high quality pelleted feed of 1.72 Mcal/kg net energy (which is assumed to be 13 kJ DE/g) by pressing a panel. Lights were on from 07:00 till 19:00 h. To imitate these experimental settings in the model, the variables sex, temperature, light period, dark period, and DE content diet were adjusted to fit the empirical data of Bigelow and Houpt (1988). The development of feeding rate in pigs in the study of Bigelow and Houpt (1988) did not follow a linear function like the equation in our model, which was computed based on mechanistic processes (equation 15 in appendix A.2). To fit the study of Bigelow and Houpt (1988), the equation for feeding rate in our model was replaced by the mean values of feeding rate observed in their empirical study (Fig. 4). Fig. 5 shows the comparison of empirically measured feeding patterns of pigs in the study of Bigelow and Houpt (1988) with the simulated feeding patterns in our model. In the empirical data, meal frequency decreased as BW increased, whereas meal duration tended to decrease (Bigelow and Houpt, 1988). Meal size and interval time between meals increased until pigs reached about 80 kg BW, whereas daily feeding time seemed to decrease until 80 kg

Body weight (kg) Fig. 4. Feeding rate in the model based on mean values of feeding rate per body weight category of pigs from the study of Bigelow and Houpt (1988).

BW (Bigelow and Houpt, 1988). Patterns of meal frequency, meal size, interval time between meals, and feeding time resulting from the model corresponded to empirical results until 120 kg BW. After 120 kg, meal size, meal frequency and interval time between meals showed some deviation. Meal duration in the model was relatively constant, while the empirical data of Bigelow and Houpt (1988) showed more variation in meal duration. 3.2. The effect of feeding rate on feeding patterns and growth Adjusting the feeding rate to the empirical values of Bigelow and Houpt (1988) was needed to reproduce feeding patterns reported in their study. We compared, therefore, model results based on empirical values of Bigelow and Houpt (1988) with results using the preferred feeding rate (equation 15 in appendix A.2). Fig. 6 shows that feeding rate affects meal frequency, meal size, meal duration, meal interval time and feeding time. Feed intake and BW gain did

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

Body weight (kg)

130

120

110

130

120

110

100

90

1000

130

120

110

100

90

80

70

0 60

130

120

110

100

90

80

70

60

50

0

2000

50

50

3000

40

100

4000

30

Daily feed intake (g)

150

40

80

Body weight (kg)

200

30

70

0

Body weight (kg)

Daily feeding me (min)

90

50

60

130

120

100

110

90

80

70

60

50

40

5

100

50

10

150

40

15

200

30

20

Meal interval me (min)

Body weight (kg)

25

30

Meal duraon (min)

Body weight (kg)

100

30

130

110

120

100

90

80

70

0 60

0 50

100

40

3

80

200

70

6

300

60

9

400

50

12

40

Meal size (g)

15

30

Meal frequency (no/day)

74

Body weight (kg)

Fig. 5. Validation of the model by comparing empirical feeding patterns based on feeding data (in which they excluded small feeding bouts of less than 10% of the average meal size) of six female Yorkshire pigs in experimental data of Bigelow and Houpt (1988) () with results of 6 simulation runs of one gilt with a feeding rate adjusted to the data of Bigelow and Houpt (1988) in the model (- - -).

not show a clear response. BW gain of pigs was slightly lower for empirical values compared with modelled values at the end of the simulation period. 3.3. Sensitivity analysis The sensitivity analysis involved varying the values of main parameters individually by 20% above (S+ ) or below (S− ) their reference value. Table 4 shows the parameters included and their values used in the analysis. Final BW, mean daily feed intake, mean meal size and mean meal duration were chosen to represent the effect of parameters on growth and feeding patterns. Feed intake, and to a lesser extent final BW, was mainly affected by Diet palatability, Response to light and Digestion duration. Meal size and meal duration were mainly affected by Reinforcement feeding motivation, and to a lesser extent by Diet palatability. Fig. 7 shows the effect of Diet palatability on all feeding patterns and BW gain. Decreasing Diet palatability had a limited effect on meal frequency, meal interval time, and feeding time, while increasing Diet palatability had an effect on all feeding patterns. 4. Discussion Our model aimed at understanding the development of feeding patterns and related growth of individual pigs, which required the inclusion of relatively comprehensive mechanisms. To understand how feeding behaviour is controlled, a model requires considerable detail in the processes (Tichit et al., 2009). We included variables on the level of detail that was necessary to integrate motivation, behaviour and growth. The variable energy absorption,

for example, affected both growth and motivation in the short and long term. Additionally, we expect that including this level of detail in our model is important for its future use, when more agents (i.e. more pigs) are to be included. Interactions among agents can affect processes underlying pig decisions on different levels. Competition for feed among pigs, for example, can increase the attractiveness of the feed, the feeding drive of a pig, or increase motivations other than feeding, such as avoiding fights or showing synchronised behaviours. The validation showed that model predictions for the development of feeding patterns over time were comparable to the empirical data of Bigelow and Houpt (1988), except for meal duration. Meal duration showed little variation in the model compared to the empirical study. The low variation can be explained by the parameter Reinforcement feeding motivation. This parameter was a theory-based parameter and no clear evidence for a value was available. A dynamic parameter value, based on the state of the animal (e.g. body weight or hunger level), would have been best, but such a value would have increased complexity and uncertainty in the model. Therefore, a fixed and calibrated value for Reinforcement feeding motivation was chosen. Although the fixed parameter caused a less accurate pattern of meal duration in the model, the effect on other feeding patterns and performance was limited. The sensitivity analysis showed that an increase of Reinforcement feeding motivation prolonged meal duration and increased meal size, whereas feed intake and BW were not affected. This corresponds to the suggestion of Nielsen (1999) that animals adjust their feeding patterns to reach a certain level of energy intake. Thus, when meal duration and meal size decreased, the pig increased its meal frequency in order to reach the same feed intake.

500 400 300 200 100 110

120

130

110

120

130

100

90

100

110

120

130

120

130

100

110

90

80

70

300

Body weight (kg)

100

90

80

70

0 60

130

120

110

100

90

80

0

600

50

10

900

40

20

1200

30

30

Body weight gain (g/day)

40

70

60

0

Body weight (kg)

50

60

90

1000

Body weight (kg)

50

80

2000

130

120

110

100

90

80

70

60

50

40

0

3000

50

50

4000

40

100

30

70

0

30

Daily feed intake (g)

150

40

80

50

Body weight (kg)

200

30

60

100

60

130

120

110

100

90

80

70

60

50

40

11

150

50

11.5

200

40

12

250

30

12.5

Meal interval me (min)

13

30

Meal duraon (min)

Body weight (kg)

Body weight (kg)

Daily feeding me (min)

70

50

40

30

130

120

110

90

100

80

70

60

50

40

0

Body weight (kg)

Feeding rate (g/min)

75

600

Meal size (g)

14 12 10 8 6 4 2 0 30

Meal frequency (no/day)

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

Body weight (kg)

Fig. 6. The effect on feeding patterns and body weight gain of two different types of feeding rates in the model: feeding rate based on the preferred feeding rate and affected by palatability and satiation (––––) and the mean feeding rate values as observed in the empirical study of Bigelow and Houpt (1988) (- - -). Simulation runs of one gilt were repeated 50 times for each type of feeding rate with use of the reference parameter values.

Table 4 Sensitivity analysis of final body weight, mean daily feed intake, mean meal size and mean meal duration after 120 days of simulationa . Parameter (reference value)

Feed characteristics DE content diet (14.2 kJ/g) Dietary amino acid content Lysine (11 g/kg) Dietary total protein content (132 g/kg) Diet palatability (0.85) Physiological factors Response to light (0.7) Response to darkness (0.5) Reinforcement of feeding motivation (0.4) Digestion duration (1230) Pig characteristics Minimum body lipid to body protein ratio (1) Mean protein deposition gilts (137 g/min) a

Variation final body weight (%)b

Variation mean feed intake (%)b

Variation mean meal size (%)b

Variation mean meal duration (%)b

S+

S−

S+

S−

S+

S−

S+

S−

5.5 0.0 −1.0 12.0

−3.9 −0.6 −3.2 −9.2

−5.5 0.0 −0.4 28.6

13.8 −0.5 −2.8 −18.0

2.7 0.0 −0.6 9.3

−7.3 −0.7 −3.9 −15.6

−1.6 −0.1 0.0 −3.4

0.2 0.1 0.1 4.3

10.8 1.1 0.4 −3.9

−7.6 −1.0 −0.3 7.4

26.3 1.9 0.8 −8.2

−16.5 −1.1 −0.7 17.1

−8.1 −0.7 18.4 −6.7

3.4 0.9 −18.8 4.5

−1.8 −1.3 18.7 1.3

0.1 2.9 −19.1 −2.7

0.9 6.5

−0.9 −8.6

0.9 2.8

−0.9 −4.1

1.2 4.2

−1.1 −5.8

0.0 0.0

0.1 0.1

Simulation runs of one gilt were repeated 50 times with use of reference parameter values and use of the model feeding rate equation. Parameter values were altered by 20% above their reference value (S+ ) or 20% below (S− ). Sensitivity was calculated as output change from the reference output in percentage. b

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

20

Simulaon days

110

120

90

100

120

110

100

90

80

400 200 120

110

100

90

80

70

60

50

0 40

120

110

90

100

80

70

60

50

40

30

0

600

30

10

800

20

20

1000

0

30

1200

10

40

Body weight gain (g/day)

Simulaon days

50

20

70

0

Simulaon days

10

60

1000

120

110

90

100

80

70

60

50

40

30

20

10

0

2000

100 110 120

40

3000

70 80 90

60

4000

30 40 50 60

80

5000

0 10 20

Daily feed intake (g)

100

0

50

0

Simulaon days

120

0

Daily feeding me (min)

80

50

Simulaon days

Feeding rate (g/min)

70

100

40

120

110

90

100

80

70

60

50

40

30

20

10

10.5

150

30

11

200

20

11.5

250

0

12

300

10

13

Meal interval me (min)

Simulaon days

12.5

0

Meal duraon (min)

Simulaon days

60

0

120

110

90

100

70

80

60

50

40

0 30

0 20

200

0

2

50

400

40

4

600

30

6

20

8

800

10

Meal size (g)

10

10

Meal frequency (no/day)

76

Simulaon days

Fig. 7. The effect of diet palatability on the development of mean feeding patterns and daily body weight gain. (–– Reference setting, - - - +20%, ······ −20%). Simulation runs of one gilt were repeated 50 times with use of reference parameter values and use of the model feeding rate equation.

Despite the inability to vary meal duration, the model still showed consistency with the empirical results of Bigelow and Houpt (1988) on meal frequency, meal size, interval time between meals and feeding time until the pig weighed 120 kg. The deviation in feeding patterns after 120 kg could be a result of the low variation in meal duration. Because the pig in our model could not increase its meal duration, it increased its meal frequency instead of its meal size. Model results were difficult to validate with other studies as most studies observed feeding patterns of pigs for a limited BW range and in group-housed conditions (e.g. Hyun et al., 1997), or reported an average value for feeding patterns during the growing period (e.g. De Haer and Merks, 1992). Furthermore, comparison with other studies was limited due to different methods for defining meal criteria and reporting feeding patterns, for instance in feeding bouts or visits (Maselyne et al., 2015). In accordance with Bigelow and Houpt (1988), studies on feeding patterns in pigs showed in general an increase in feed intake, meal size and feeding rate, and a decrease or varying trend in meal frequency, meal duration, and

feeding time when pigs gained weight (e.g. Nienaber et al., 1990; Walker, 1991; Hyun et al., 1997; Fàbrega et al., 2003). To be able to reproduce feeding patterns for validation, it was important to have a similar feeding rate in our model to what was observed in the empirical study of Bigelow and Houpt (1988). The feeding rate equation in our model produced a linear increase of feeding rate, similar to empirical studies (e.g. Hyun et al., 1997; Rauw et al., 2006). The feeding rate in the study of Bigelow and Houpt (1988), however, was not increasing linearly, especially after pigs reaching approximately 80 kg of BW. An explanation for the change in feeding rate is not given in that study, but it could be related to conditions in the study such as the reported variations in water intake or the required panel pressing to obtain feed. The effect of feeding rate on other feeding patterns in the model showed similar inter-relatedness of feeding patterns as reviewed by Nielsen (1999): when meal frequency, meal size and meal duration are known, daily feed intake, daily feeding time and feeding rate can be derived from these parameters. Although feeding rate affected

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

most feeding patterns, it did not affect daily feed intake as pigs adapted their feeding patterns to reach the same feed intake level. For future research, it would be interesting to study the development of feeding rate for group-housed pigs in the model. Changes in feeding rate of group-housed pigs are not fully understood and could be an interesting indicator for social pressure (Nielsen, 1999). Feeding rate, meal duration or other feeding patterns did not affect feed intake and BW gain. There, however, was a small deviation in BW gain of pigs when using the feeding rate values of Bigelow and Houpt (1988) at the end of the simulations. Due to the reduced feeding rate, pigs needed to increase their feeding time to reach the same daily feed intake as with a higher feeding rate. This increase in feeding time resulted in more energy use per day, and consequently a slightly decreased BW gain. The sensitivity analysis showed that Diet palatability was the parameter affecting meal size, feed intake and BW gain most. The effect of Diet palatability on feed intake corresponds to empirical results. Several studies in humans showed the effect of increased food intake and larger meal sizes with more palatable food (see for review Sørensen et al., 2003). Studies in pigs showed increased feed intake and daily BW gain with preferred diets (e.g. Janz et al., 2007). Besides Diet palatability, Response to light also affected feed intake and BW gain. This could indicate that light intensity can affect feeding behaviour. There is no indication in literature, however, that feed intake and BW gain in pigs are affected by the intensity of light (e.g. Wheelhouse and Hacker, 1982). A more likely explanation is that response to light is an individual pig trait. Individual differences among pigs were shown, for example, in melatonin response to darkness (Tast et al., 2001). Melatonin response is related to the sleep-wakefulness cycle, which follows the same diurnal rhythm as feeding behaviour, and in which sleeping and active behaviours alternate (Reilly and Waterhouse, 2007). A large individual variation in response to light, however, is not expected, as different light intensities did not affect melatonin response in pigs above a minimum level of 40 lx light intensity (Tast et al., 2001). Due to the modelled conditions of ad libitum feed access and no social competition, we assumed that behavioural decision-making was not affected by emotions. This might be partly besides the truth as studies showed that social isolation can be stressful for individually housed pigs (e.g. Herskin and Jensen, 2000; Ruis et al., 2001) and can affect feeding motivation (Pedersen et al., 2002). Where Nielsen (1999) suggested that pigs in a barren environment without social contact can over-consume feed, Pedersen et al. (2002) showed that pigs valued food more in the company of another pig. These effects were considered secondary in our current model and not expected to change feeding patterns considerably. This study was the first step in understanding feeding behaviour in pigs. The model presented here showed that combining growth and motivation can explain the changes in the feeding patterns of an individually housed growing pig over time, mostly representing internal factors. Growing pigs, however, are usually housed in groups. Group-housed pigs are known to exhibit different feeding patterns than individually housed pigs, as they feed with a higher feeding rate and have larger, but less frequent, meals (De Haer and Merks, 1992). In addition, group-housed pigs showed a lower feed intake (De Haer and Merks, 1992) and productivity level (Gonyou et al., 1992; De Haer and De Vries, 1993b) compared with individually housed pigs. These differences in feeding behaviour and productivity can be due to social factors, such as social facilitation, competition and stress (Bornett et al., 2000). Additionally, individual differences in the alteration of feeding patterns were observed among group-housed pigs that were not understood (Bornett et al., 2000). The current model serves as a basis to include these individual differences and social factors, and to explore the role of social interaction in the causation of feeding patterns in group-housed pigs.

77

5. Conclusion This modelling study provided insight in how growth and motivation explain the development of feeding patterns of an individually housed pig over time. Pig and feed characteristics affected the motivation to reach a desired level of daily feed intake. Without feeding restrictions, pigs adapted feeding patterns to reach this daily feed intake without affecting feed intake and growth. The use of agent-based modelling for understanding pig behaviour is a novel and promising approach. The developed model is suitable to further study mechanisms underlying feeding behaviour and performance of group-housed pigs. Conflicts of interest The authors declare no conflict of interest. Acknowledgements We would like to thank Walter Gerrits for comments on an earlier version of the model and Aart van der Linden for advice on modelling. Furthermore, we would like to thank Laura Webb for commenting on this manuscript. This research was funded by the IP/OP program ‘Complex Adaptive Systems’ of Wageningen UR. Appendix A. Technical details A.1. Details of the sub-model ‘Motivational decision-making’ An animal is motivated to feed if its feeding drive (F drivem ) is higher than its satiation (F satiationm ) expressed per minute. F drivem is a function based on the energy balance of the animal (E balanced ) each minute, palatability of the feed (Diet palatability) and the diurnal rhythm (D rhythm) (Sauvant et al., 1996). For pigs it is known that the nutrient balance can affect feeding behaviour (NRC, 2012) and, therefore, the nutrient balance (N balance) was also added to Equation 1. As the pig in the model received ad libitum feed that met its nutritional requirements, it is assumed that N balance had no effect on F drivem . In that case, the value for N balance was set to 1.0. F drivem = E balanced × Diet palatability × D rhythm × N balance (1) For Diet palatability a fixed dimensionless value (0.85) was used because it was assumed that feed is always available with the same potential palatability. The value 0.85 was chosen after calibration of the model on expected amount of feed intake. D rhythm was represented by two dimensionless values: a higher value for the response to light (Response to light) from 08:00 to 19:00 hour (0.7) and a lower value for the response to darkness (Response to darkness) from 19:00 to 08:00 hour (0.5). The values 0.7 and 0.5 were chosen after calibration of the model on expected distribution of feed intake over a 24 h period. E balanced was included in the model as a coefficient equal to 1 if the digestible energy (DE) absorption (kJ DE) of a pig that day (E absorptiond ) was high enough to meet the energy requirements (kJ DE) of that day (E requirementd ). If the energy requirements were not met, the coefficient was higher than 1 and increased the feeding drive. E balanced was calculated analogously to the model of Sauvant et al. (1996). E balanced = Maximum

 E requirement d E absorptiond



,1

(2)

The equation for E absorptiond was adjusted to make a better fit to a situation of an intensively, individually housed pig and the

78

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

parameters in this model. E absorptiond (kJ DE) was calculated as the sum of absorbed energy (E absorptionm ), based on passed minutes that day (m) with a maximum of 1440. E absorptiond =

m 

Gut loadm = exp E absorptionm

(3)

m=1

E absorptionm (kj DE) was the amount of energy that was absorbed in the gut each minute based on digested feed (F digestedm ) and the energy content of the diet (Diet DE, kJ DE/g). E absorptionm = F digestedm × Diet DE

(4)

F digestedm (g) was the amount of feed that was digested in the gut each minute based on the gut content (Gut contentm ) and duration of digestion (Diet digestion). F digestedm = Gut contentm × 1000 ×

1 Diet digestion

(5)

In a study of Lewis and McGlone (2008) on pigs with ad libitum feed, the duration of digestion was on average 20.5 (+/- 0.31) h, with a range of 18 to 24 h. Therefore, it was assumed in the model that Diet digestion was 20.5 hours (1230 minutes). E requirementd (kJ DE) was based on energy use for maintenance and activity and growth capacity since the start of that day. Total energy use was calculated as the sum (with a maximum of 1440 min) of maintenance (E maintenancem , Equation 16) and activity (E activitym , Equation 19) during passed minutes that day. The potential growth capacity for that day (G capacityd ) was also based on passed minutes that day. E requirementd =

+

 G capacity d 1440

m  m=1

E maintenancem +



m 

×m

(6)

G capacityd = E intaked − PrevE maintenanced − PrevE activityd (7) E intaked (kJ DE) was calculated from sex specific equations as described in NRC (2012), converted to DE in kJ. Effective diet metabolisable energy (ME) content of growing pigs was converted to DE by multiplying with 1.03 (NRC, 2012), and converted from kcal to kJ by multiplying with 4.187. F satiationm was calculated in a similar way as in the model of Sauvant et al. (1996), where it was based on the rumen load and the instantaneous energy balance (E balancem ). Rumen load was based on gut load (Gut loadm ). F satiationm = Gut loadm × exp(0.1×E

balancem )

(8)

E balancem was calculated based on E absorptionm and the instantaneous requirement of energy (E requirementm ) per minute. E absorptionm − E requirementm = E requirementm

(9)

With E requirementm (kJ DE) based on the daily energy requirements per day. E requirementm =

E intaked 1440

Gut contentm = Gut contentm−1 −

(10)

(11)

 F digested  m 1000

(12)

A.2. Details of feeding rate determination Feeding rate of pigs is related to mouth capacity (Illius and Gordon, 1987), which depends on the size of the animal (Nienaber et al., 1990; Nielsen, 1999) and increases with age and BW (Rauw et al., 2006). Furthermore, feeding rate is related to feed composition (Brouns et al., 1994; Brouns et al., 1997). Wellock et al. (2003) included these effects in an equation to calculate the maximum feeding rate of growing pigs (FR max, g per minute). A value of 2.85 was assumed for a high quality pelleted feed and 3.6 for the water-holding capacity related to that feed (Wellock et al., 2003). FR max =

m=1

Gut contentm −Gut sizem Gut sizem

Gut sizem (kg) was calculated each minute based on the equation of NRC (2012) to calculate gut fill estimated from empty body weight (EBWm ). In the model of Sauvant et al. (1996), gut content was based on ruminal DM and DM content (also based on the size of the feed particles). Pigs, however, are monogastric animals and have no rumen. Gut contentm (kg) in this study was calculated by reducing the amount of feed in the gut in the previous minute (Gut contentm-1 , kg) with F digestedm each minute. At start of the model, the initial value of Gut contentm-1 was set to a random value based on a normal distribution with a mean of 0.5 × Gut sizem and standard-deviation of 0.05 × Gut sizem .



E activitym

G capacityd (kJ DE) was based on the default daily DE intake of pigs (E intaked ) reduced by the sum of energy costs of the previous day for body maintenance (PrevE maintenanced ) and activity (PrevE activityd ). At the start of the model, the initial sum of energy costs for PrevE maintenanced and PrevE activityd is set to 5800, estimated for a pig with 26.5 kg BW and 18% of the time activity costs.

E balancem

Gut loadm was a function calculated each minute and based on gut size (Gut sizem ) and Gut contentm , following the equation for rumen load proposed by Sauvant et al. (1996).

2.85 × BW 1.0



3.6

(13)

FR max is comparable to feeding rates of pigs housed in groups of 30 in the study of Walker (1991), who investigated the effect of feeder availability on feeding behaviour. Wellock et al. (2003) assumed that the feeding rates of these pigs represented maximum feeding rates, because the increase in group size above 20 pigs (and thus increasing feeder occupation) showed a limited further increase in feeding rate. Feeding rates of pigs around 80 kg of BW housed in groups of 10 were almost halved compared to groups of 30 pigs (Walker, 1991). In addition, feeding rates of individually housed pigs were lower than feeding rates of pigs in groups of 8 (De Haer and Merks, 1992). Nielsen (1999) suggested that pigs have a preferred feeding rate, which can increase due to social effects. Based on the work done by Bigelow and Houpt (1988) and Nienaber et al. (1990), we assumed that observed feeding rates of individually housed and ad libitum fed pigs could represent preferred feeding rates. Therefore, preferred feeding rate (FR preferred, g per minute) was calculated to fit in the range of these feeding rates. FR preferred = FR max ×BW −0.25

(14)

Additional factors that can affect feeding rate are feed palatability and the level of satiation (Sauvant et al., 1996). Therefore these factors were included in the equation to calculate the actual feeding rate in g per minute (FRm ). FRm =

FR preferred × Diet palatability F satiationm

(15)

A.3. Details of the sub-model Growth In the model of de Lange (1995), energy requirements for maintenance and activity were calculated by one equation as a function of metabolic BW. To include the effect of temperature and variability of animal activity in the model, equations from NRC (2012)

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

were converted to DE in kJ per minute and included in the model to calculate utilisation of energy for maintenance (E maintenancem , kJ DE), based on energy use for body maintenance (E bodym ) and energy use for thermogenesis (E thermogenesism ). E maintenancem = E bodym + E thermogenesism

(16)

E bodym (kJ DE) was based on BW (NRC, 2012).



E bodym =

197 × BW 0.6 1440



× 1.03 × 4.187

(17)

E thermogenesism (kJ DE) was based on lower critical temperature (LCT), temperature (T) and E bodym (NRC, 2012). E thermogenesism = 0.07425 × (LCT − T ) × E bodym

(18)

When a pig is resting or lying energy use is calculated as maintenance energy costs. When a pig is feeding, drinking, exploring or standing additional energy costs for activity are added. Energy costs for activity per minute (E activitym , kJ DE) were based on the equation of Van Milgen et al. (1998) and were dependent on the muscle-mass (Muscle massm ) of the body. E activitym =

21 × Muscle mass0.91 × 1.03 60

(19)

Muscle massm (kg) was based on an estimated percentage of muscle mass in Large White castrated males of 43% of EBWm (Van Milgen et al., 1998). Potential body protein growth was based on protein deposition (PD) in a pig. In the model of de Lange (1995), PD was determined by the lowest value of two factors: protein intake and the fixed value of maximum protein retention of a pig per day. To include specific BW effects on PD, equations from NRC (2012) were included, which defined PD (g/day) curves for pigs between 25 and 125 kg BW, based on a mean PD value (PD mean) and a standard sex-specific equation. To convert PD in g/day to g/minute, the value was divided by 1440. BW and body composition was based on EBWm and gut fill (NRC, 2012). EBWm was calculated analogously to NRC (2012) by the sum of whole-body lipid mass, whole-body protein mass, wholebody water mass and whole-body ash mass. Gut fill (Gut sizem ) is calculated from EBWm or initial BW (NRC, 2012). References Asher, L., Collins, L.M., Ortiz-Pelaez, A., Drewe, J.A., Nicol, C.J., Pfeiffer, D.U., 2009. Recent advances in the analysis of behavioural organization and interpretation as indicators of animal welfare. J. R. Soc. Interface 6, 1103–1119. Batterham, E.S., Bayley, H.S., 1989. Effect of frequency of feeding of diets containing free or protein-bound lysine on the oxidation of [C]lysine or [C]phenylalanine by growing pigs. Br. J. Nutr. 62, 647–655. Beattie, V.E., O’Connell, N.E., Moss, B.W., 2000. Influence of environmental enrichment on the behaviour, performance and meat quality of domestic pigs. Livest. Prod. Sci. 65, 71–79. Berthoud, H.-R., Morrison, C., 2008. The brain, appetite, and obesity. Annu. Rev. Psychol. 59, 55–92. Bigelow, J.A., Houpt, T.R., 1988. Feeding and drinking patterns in young pigs. Physiol. Behav. 43, 99–109. Black, J.L., 1995. Modelling energy metabolism in the pig - critical evaluation of a simple reference model. In: Moughan, P.J., Verstegen, M.W.A., Visser-Reyneveld, M.I. (Eds.), Modelling Growth in the Pig. Wageningen Pers, Wageningen, pp. 87–102. Black, J.L., Lange, C.F.M.d., 1995. Introduction to the principles of nutrient partitioning for growth. In: Moughan, P.J., Verstegen, M.W.A., Visser-Reyneveld, M.I. (Eds.), Modelling Growth in the Pig. Wageningen Pers, Wageningen, pp. 33–45. Bolhuis, J.E., Schouten, W.G.P., Schrama, J.W., Wiegant, V.M., 2005. Behavioural development of pigs with different coping characteristics in barren and substrate-enriched housing conditions. Appl. Anim. Behav. Sci. 93, 213–228. Bornett, H.L.I., Morgan, C.A., Lawrence, A.B., Mann, J., 2000. The effect of group housing on feeding patterns and social behaviour of previously individually housed growing pigs. Appl. Anim. Behav. Sci. 70, 127–141. Brouns, F., Edwards, S.A., English, P.R., 1994. Effect of dietary fibre and feeding system on activity and oral behaviour of group housed gilts. Appl. Anim. Behav. Sci. 39, 215–223.

79

Brouns, F., Edwards, S.A., English, P.R., 1997. The effect of dietary inclusion of sugar-beet pulp on the feeding behaviour of dry sows. Anim. Sci. 65, 129–133. Day, J.E., Kyriazakis, I., Rogers, P.J., 1998. Food choice and intake: towards a unifying framework of learning and feeding motivation. Nutr. Res. Rev. 11, 25–44. Day, J.E.L., Kyriazakis, I., Lawrence, A.B., 1995. The effect of food deprivation on the expression of foraging and exploratory behaviour in the growing pig. Appl. Anim. Behav. Sci. 42, 193–206. De Haer, L.C.M., De Vries, A.G., 1993a. Effects of genotype and sex on the feed intake pattern of group housed growing pigs. Livest. Prod. Sci. 36, 223–232. De Haer, L.C.M., De Vries, A.G., 1993b. Feed intake patterns of and feed digestibility in growing pigs housed individually or in groups. Livest. Prod. Sci. 33, 277–292. De Haer, L.C.M., Merks, J.W.M., 1992. Patterns of daily food intake in growing pigs. Anim. Sci. 54, 95–104. De Lange, C.F.M., 1995. Framework for a simplified model to demonstrate principles of nutrient partitioning for growth in the pig. In: Moughan, P.J., Verstegen, M.W.A., Visser-Reyneveld, M.I. (Eds.), Modelling Growth in the Pig. Wageningen Pers, Wageningen, Netherlands, pp. 71–85. De Lange, C.F.M., Morel, P.C.H., Birkett, S.H., 2003. Modeling chemical and physical body composition of the growing pig. J. Anim. Sci. 81, E159–E165. Fàbrega, E., Tibau, J., Soler, J., Fernández, J., Font, J., Carrión, D., Diestre, A., Manteca, X., 2003. Feeding patterns, growth performance and carcass traits in group-housed growing-finishing pigs: the effect of terminal sire line, halothane genotype and age. Anim. Sci. 77, 11–22. Gonyou, H.W., Chapple, R.P., Frank, G.R., 1992. Productivity, time budgets and social aspects of eating in pigs penned in groups of five or individually. Appl. Anim. Behav. Sci. 34, 291–301. Herskin, M.S., Jensen, K.H., 2000. Effects of different degrees of social isolation on the behaviour of weaned piglets kept for experimental purposes. Anim. Welf. 9, 237–249. Hogan, J.A., 1997. Energy models of motivation: A reconsideration. Appl. Anim. Behav. Sci. 53, 89–105. Hyun, Y., Ellis, M., McKeith, F.K., Wilson, E.R., 1997. Feed intake pattern of group-housed growing-finishing pigs monitored using a computerized feed intake recording system. J. Anim. Sci. 75, 1443–1451. Illius, A.W., Gordon, I.J., 1987. The allometry of food Intake in grazing ruminants. J. Anim. Ecol. 56, 989–999. Janz, J.A.M., Morel, P.C.H., Wilkinson, B.H.P., Purchas, R.W., 2007. Preliminary investigation of the effects of low-level dietary inclusion of fragrant essential oils and oleoresins on pig performance and pork quality. Meat Sci. 75, 350–355. Johnson, A.W., 2013. Eating beyond metabolic need: how environmental cues influence feeding behavior. Trends Neurosci. 36, 101–109. Kanis, E., 1990. Effect of food intake capacity on production traits in growing pigs with restricted feeding. Anim. Sci. 50, 333–341. Lewis, C.R.G., McGlone, J.J., 2008. Modelling feeding behaviour, rate of feed passage and daily feeding cycles, as possible causes of fatigued pigs. Animal 2, 600–605. Maselyne, J., Saeys, W., Van Nuffel, A., 2015. Review: Quantifying animal feeding behaviour with a focus on pigs. Physiol Behav 138, 37–51. Mason, G., Bateson, M., 2009. Motivation and the organization of behaviour. The Ethology of Domesticated Animals, 2nd edn. CAB International, Wallingford, UK, pp. 38–56. McFarland, D.J., Sibly, R.M., 1975. The behavioural final common path. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 270, 265–293. Morgan, C.A., Emmans, G.C., Tolkamp, B.J., Kyriazakis, I., 2000. Analysis of the feeding behavior of pigs using different models. Physiol. Behav. 68, 395–403. Moughan, P.J., 1995. Modelling protein metabolism in the pig - first principles. In: Moughan, P.J., Verstegen, M.W.A., Visser-Reyneveld, M.I. (Eds.), Modelling Growth in the Pig. Wageningen Pers, Wageningen, pp. 59–70. Nielsen, B.L., 1999. On the interpretation of feeding behaviour measures and the use of feeding rate as an indicator of social constraint. Appl. Anim. Behav. Sci. 63, 79–91. Nienaber, J.A., McDonald, T.P., Hahn, G.L., Chen, Y.R., 1990. Eating dynamics of growing-finishing swine. Trans. ASAE 33, 2011–2018. NRC, 2012. Nutrient Requirements of Swine. National Academies Press, Washington, DC. Nyachoti, C.M., Zijlstra, R.T., De Lange, C.F.M., Patience, J.F., 2004. Voluntary feed intake in growing-finishing pigs: A review of the main determining factors and potential approaches for accurate predictions. Can. J. Anim. Sci. 84, 549–566. Pedersen, L.J., Jensen, M.B., Hansen, S.W., Munksgaard, L., Ladewig, J., Matthews, L., 2002. Social isolation affects the motivation to work for food and straw in pigs as measured by operant conditioning techniques. Appl. Anim. Behav. Sci. 77, 295–309. Quiniou, N., Dubois, S., Noblet, J., 2000. Voluntary feed intake and feeding behaviour of group-housed growing pigs are affected by ambient temperature and body weight. Livest. Prod. Sci. 63, 245–253. Railsback, S.F., Grimm, V., 2012. Agent-Based and Individual-Based Modeling: A Practical Introduction. Princeton University Press, Princeton, New Jersey. Rauw, W.M., Soler, J., Tibau, J., Reixach, J., Gomez Raya, L., 2006. Feeding time and feeding rate and its relationship with feed intake, feed efficiency, growth rate, and rate of fat deposition in growing Duroc barrows. J. Anim. Sci. 84, 3404–3409. Reilly, T., Waterhouse, J., 2007. Altered sleep–wake cycles and food intake: The Ramadan model. Physiol. Behav. 90, 219–228. Ruis, M.A.W., te Brake, J.H.A., Engel, B., Buist, W.G., Blokhuis, H.J., Koolhaas, J.M., 2001. Adaptation to social isolation: Acute and long-term stress responses of growing gilts with different coping characteristics. Physiol. Behav. 73, 541–551.

80

I.J.M.M. Boumans et al. / Applied Animal Behaviour Science 171 (2015) 69–80

Sauvant, D., Baumont, R., Faverdin, P., 1996. Development of a mechanistic model of intake and chewing activities of sheep. J. Anim. Sci. 74, 2785–2802. Sørensen, L.B., Møller, P., Flint, A., Martens, M., Raben, A., 2003. Effect of sensory perception of foods on appetite and food intake: a review of studies on humans. Int. J. Obes. 27, 1152–1166. Tast, A., Love, R.J., Evans, G., Andersson, H., Peltoniemi, O.A.T., Kennaway, D.J., 2001. The photophase light intensity does not affect the scotophase melatonin response in the domestic pig. Anim. Reprod. Sci. 65, 283–290. Thingnes, S.L., Ekker, A.S., Gaustad, A.H., Framstad, T., 2012. Ad libitum versus step-up feeding during late lactation: The effect on feed consumption, body composition and production performance in dry fed loose housed sows. Livest. Sci. 149, 250–259. Tichit, M., Puillet, L., Sauvant, D., 2009. How models contribute to livestock farming system research: overview of recent advances. Options Méd., 293–306. Van Milgen, J., Bernier, J.F., Lecozler, Y., Dubois, S., Noblet, J., 1998. Major determinants of fasting heat production and energetic cost of activity in growing pigs of different body weight and breed/castration combination. Br. J. Nutr. 79, 509–517.

Walker, N., 1991. The effects on performance and behaviour of number of growing pigs per mono-place feeder. Anim. Feed Sci. Technol. 35, 3–13. Wellock, I.J., Emmans, G.C., Kyriazakis, I., 2003. Predicting the consequences of social stressors on pig food intake and performance. J. Anim. Sci. 81, 2995–3007. Wellock, I.J., Emmans, G.C., Kyriazakis, I., 2004. Modeling the effects of stressors on the performance of populations of pigs. J. Anim. Sci. 82, 2442–2450. Wheelhouse, R.K., Hacker, R.R., 1982. The effect of four different types of fluorescent light on growth, reproductive performance, pineal weight and retinal morphology of Yorkshire gilts. Can. J. Anim. Sci. 62, 417–424. Wilensky, U., 1999. Netlogo, Center for Connected Learning and Computer-Based Modeling. Northwestern University, Evaston, Illinois. Williams, N.H., Stahly, T.S., Zimmerman, D.R., 1997. Effect of level of chronic immune system activation on the growth and dietary lysine needs of pigs fed from 6 to 112 kg. J. Anim. Sci. 75, 2481–2496. Yoosuk, S., Ong, H.B., Roan, S.W., Morgan, C.A., Whittemore, C.T., 2011. A simulation model for predicting the voluntary feed intake of a growing pig. Acta Agric. Scand. Sect. A Anim. Sci. 61, 168–186.